Community Fibre performance

So Community Fibre finally went live after some delays and I've had the service installed. I'll update my other thread with details of the installation method, but for now here are some performance figures. In short, it seems to deliver as advertised.

Location: SW London, overhead/PIA served house. CF build completed July 2021.
Product: 1000/1000

Latency at 10am on a Thursday - min,average,max,std.dev:
bbc.co.uk (151.101.192.81) 1.7,1.8,2.2,0.1 - via Telehouse North it seems
1.1.1.1 (CF issue as secondary DNS) 2.6,2.7,3.0,0.1 - via THN / LINX LON1
9.9.9.9 (which is the primary DHCP DNS) 1.6,1.8,2.2,0.1 - also via THN
8.8.8.8 2.4,2.6,2.6,0.1
18.130.0.0 (AWS eu-west-2) 3.3,3.4,3.5,0.1
54.90.0.3 (AWS us-east-1) 77.6,77.7,77.8,0.1 - NTT on the transatlantic
A VPS in OVH London (Erith) 2.1,2.3,5.5, 0.2
No major change to the above at 2330 on a Friday.

TBB graph - the (small) spikes are the Samknows box testing.

Speed has been fairly consistently above 900 up and down using Speedtest.net to London servers. fast.com tells a similar story. TBB single thread is significantly lower, around 250Mbit.

The Samknows Whitebox reports a fairly consistent 940Mbit, which regularly drops to 910 at peak time. On one occasion 800 down was recorded in the evening peak. It has only been running a few days. Upload is a flat line at 940Mbit. My area does not yet have the 3Gbit package available - I have no need of it, but it's perhaps an indication that they're operating on limited backhaul in this early stage of deployment.

Interestingly when selecting speedtest.net servers further away (Manchester, Edinburgh, Dublin, Amsterdam) many are well connected enough to achieve >800Mbit download, but the upload always seems to max out around 500. I'm not sure why that might be.

It's straightforward to connect your own router - I've cloned the MAC address of the supplied router because it was easy and there's a report somewhere online of issues with requesting DHCP leases from multiple MAC addresses in a short period of time. The v4 lease time is 24h and the IP seems sticky enough.

IPv6 seems straightforward too - even if you request a /64 prefix you get a /48 on an 8 day lease. I've only carried out basic testing but can't see any issues.

In short there are very few reasons not to go for it if it's available. A static IP would be nice but I think they're trying to use this to differentiate the business service which is considerably more expensive. For me, between dynamic DNS and VPN'ing some static IPs from a datacentre I can cope with the dynamic IP.

I'll report back after a few months when more subscribers have signed up and load is higher. For now it's likely that I'm the only one on my PON.

A new thread today made me think of both your Community Fibre threads, and I'd meant to reply to this one but got side tracked and forgot!

How are you getting on with the service? Did you ever get to the bottom of the difference you saw with TBB single-thread DL performance vs multi-thread?

Otherwise hope its all going well and living up to expectations.

I had mine installed on the 9th July worked great on the supplied router, but wanted to use my own router like I do with BT.

I had that exact same issue where the router's DHCP Client was failing to connect for both IPv4 and IPv6, so I contacted them via Twitter and was given a number to phone and spoke to one of their Tech people who could see my Routers connection and its MAC and cleared the old ones and now it connects fine.

I am from east London about 4 or so miles from Telehouse, I just tried a speed on a Manchester Server and I got this: >> Result <<

Also going by my SamKnows White Box I am getting 940Mbps up and down with a minimum of 930Mbits 24/7 so very happy, works out cheaper than what we are paying for BT.

I also can confirm IPv4 (/25 so a 18 address and 126 being usable) has a 24 hour lease time and is in fact sticky, I can do a release and get a new IP if I wanted to.

I can also confirm IPv6 has a lease time of 8days but is very sticky, even release is having issues getting a new IP.

Paul

I'm also in East London (Newham for me) and on Gigafast 1000/1000. Do you notice packet loss on the v6 side of things? I'm having a very odd issue where v4 is almost spotless in terms of packet loss, but v6 has constant loss.

This is IPv4 where the red loss is me fiddling with my router, and this is v6...

If I use something like Google Meet, which defaults to v6 when available, the thing is a stuttery mess. pfSense is also giving me packet loss warnings too. So far, have had silence from CF after sending Pingplotter data suggesting it's something going awry on their side.

A traceroute from outside your network would probably be more useful. If it's on the ISP's side you should notice packet loss on the router the hop before yours that's carried forward to your router. If their router shows no packet loss but your router does then that suggests the problem is with your router.

A traceroute to a device inside your network would probably be useful too. If you see loss on your ISP's router (the hop before yours) and loss on the pfSense router but no loss on the device inside your network then the loss isn't a problem, it's just the router prioritising other things higher than ICMPv6.

Thanks - super useful. I'll bug someone at work to run a trace from their side to mine and send over the results to observe.

Granted, IPv6 isn't super high on the priority list but nice to have so not complaining as the service is rock solid. https://ibb.co/6mrRvv4 is what I have sent CF already alongside another example showing identical behavior - it seems that loss starts occurring at hop 2 (first hop outside my network) and then throughout *.as201838.net (CFs routers). Was curious if I was the only one with this or I'd screwed something up on my side.

In reply to a post by branflakes:

I'm also in East London (Newham for me) and on Gigafast 1000/1000. Do you notice packet loss on the v6 side of things? I'm having a very odd issue where v4 is almost spotless in terms of packet loss, but v6 has constant loss.

Same, I am in Newham, I am in Manor Park to be exact.
I have the exact same issue with IPv6 when doing teh BQM, I don't think I am getting packet loss when using the actual connection via IPv6.

In reply to a post by branflakes:

This is IPv4 where the red loss is me fiddling with my router, and this is v6...

These are my current BQM's BQM Image (DDNS domain), BQM Image (IPv4), BQM Image (IPv6)
Basically the DDNS one is setup to only have IPv4, which is why it matches the IPv4 one and then there is IPv6 that is showing the packet loss.
I did a traceroute from my router to TBB BQM, to my current BT Connection, to our Rack Server along with from my current BT Connection, Rack Server and TBB back to my CFL connection and I am seeing loads of drops via IPv6, where as IPv4 is fine.

I did at first think it was my firewall on my router, but I was still having all the drops when the firewall was temp removed.

In reply to a post by branflakes:

If I use something like Google Meet, which defaults to v6 when available, the thing is a stuttery mess. pfSense is also giving me packet loss warnings too. So far, have had silence from CF after sending Pingplotter data suggesting it's something going awry on their side.

I was due a phone call a day or so ago back which they have failed to do, so I will be calling them over the next few days.

Paul

In reply to a post by aidanh:

A traceroute from outside your network would probably be more useful. If it's on the ISP's side you should notice packet loss on the router the hop before yours that's carried forward to your router. If their router shows no packet loss but your router does then that suggests the problem is with your router.

A traceroute to a device inside your network would probably be useful too. If you see loss on your ISP's router (the hop before yours) and loss on the pfSense router but no loss on the device inside your network then the loss isn't a problem, it's just the router prioritising other things higher than ICMPv6.

I am seeing the exact same issue over IPv6, here is what I got routing to TBB BQM Service last week:

/tool traceroute tool traceroute ipv6.pingbox1.thinkbroadband.com (2a02:68:1::164)
 # ADDRESS                                          LOSS SENT    LAST     AVG    BEST   WORST STD-DEV STATUS
 1 2a02:####:#:##::#                                93..  170 timeout       1     0.8     1.2     0.1
 2 lag-6.acc1.med.lon.network.as201838.net          90..  170 timeout     2.8     1.2    14.7     3.3
 3 2a02:6b60:0:1:1::59                              91..  169 timeout     3.7       1      26     6.1
 4 xge-0-0-0-25.edge1.ld8.lon.network.as201838.net  96..  169 timeout     1.9     1.8       2     0.1
 5 xge-0-0-0-5.edge1.thn.lon.network.as201838.net   89..  169 timeout     1.7     1.1     2.1     0.2
 6 2a02:6b60:0:1:2::45                              1.2%  167   1.5ms     5.4     0.7   121.4    18.8
 7 2a02:68:0:13::4                                  1.8%  167   0.7ms     3.5     0.7   149.8    14.7
 8 po4-31.core-rs4.thdo.ncuk.net                    0.6%  167   0.9ms     1.3     0.8      20     1.7
 9 ipv6.pingbox1.thinkbroadband.com                 4.2%  167   1.3ms     1.4     0.6     4.9     0.4

This is today:

/tool traceroute tool traceroute ipv6.pingbox1.thinkbroadband.com (2a02:68:1::164)
 # ADDRESS                                          LOSS SENT    LAST     AVG    BEST   WORST STD-DEV STATUS
 1 2a02:####:#:##::#                                1.7%  121   0.9ms       1     0.8     1.9     0.2
 2 lag-6.acc1.med.lon.network.as201838.net          5.8%  121   1.9ms     1.8     1.6     2.2     0.1
 3 2a02:6b60:0:1:1::59                              47..  121 timeout     1.9     1.7     2.3     0.1
 4 xge-0-0-0-25.edge1.ld8.lon.network.as201838.net   85%  120 timeout     1.8     1.7     2.1     0.1
 5 xge-0-0-0-5.edge1.thn.lon.network.as201838.net   80..  120 timeout     1.8     1.7       2     0.1
 6 2a02:6b60:0:1:2::45                              1.7%  120   1.6ms       4     1.4   126.1    14.7
 7 2a02:68:0:13::4                                  1.7%  120   0.9ms     3.7     0.7   130.6    17.7
 8 po4-31.core-rs4.thdo.ncuk.net                      0%  120  12.4ms     1.4     0.8    12.4     1.6
 9 ipv6.pingbox1.thinkbroadband.com                 2.5%  120   1.6ms     1.4     1.3     1.6     0.1

Hop 1 is my routers Gateway, as you can see there is a lot of loss, but that could be due to them prioritizing packets.
Todays routing back to our connection from TBB its a lot better:

#	Host							Sent	Recvd	Best	Avg	Worst
1 	ffff.tbb.core-rs4.thdo.ncuk.net				15	15	0ms	0ms	1ms
2 	po4-31.bdr-rt3.thdo.ncuk.net				15	15	0ms	2ms	30ms
3 	2a02:68:0:13::7						15	15	0ms	4ms	53ms
4 	lag-11.edge1.thn.lon.network.as201838.net		1	15	1ms	1ms	1ms
5 	xge-0-0-0-0.edge1.ld8.lon.network.as201838.net		4	15	1ms	1ms	1ms
6 	xge-0-0-1-0.edge-rr1.ld8.lon.network.as201838.net	11	15	1ms	1ms	1ms
7 	lag-1.acc1.med.lon.network.as201838.net			15	15	1ms	1ms	1ms
8 	lag-4.acc1.nw17.lon.network.as201838.net		15	15	1ms	1ms	1ms
9 	2a02:#:#:#:#:#:#:#					15	15	1ms	1ms	2ms

I haven needed to use IPv6 as yet so I haven't noticed any issue, I will try and use some time this week and see if anything happens.

Paul

In reply to a post by PaulKirby:

Same, I am in Newham, I am in Manor Park to be exact.
I have the exact same issue with IPv6 when doing teh BQM, I don't think I am getting packet loss when using the actual connection via IPv6.

I'm also Manor Park - E12 6 - it's a small world!

Thanks for confirming, I was worried it was just my connection but seems like it's not. My Google Meet experience could've been a bad VPN so will review again now sans VPN. I ended up switching v6 off my main LAN subnet but keeping it on the lab subnet so all the homelab kit has functioning v6 connectivity. Work has kept me far too busy so haven't extensively tested on that as of yet.

Do let us know what CF say!

Crikey you're probably neighbours, just shout out the window! 😀

In reply to a post by branflakes:

In reply to a post by PaulKirby:

Same, I am in Newham, I am in Manor Park to be exact.
I have the exact same issue with IPv6 when doing the BQM, I don't think I am getting packet loss when using the actual connection via IPv6.

I'm also Manor Park - E12 6 - it's a small world!

I am E12 5 here and yes it is LOL.

In reply to a post by branflakes:

Thanks for confirming, I was worried it was just my connection but seems like it's not. My Google Meet experience could've been a bad VPN so will review again now sans VPN. I ended up switching v6 off my main LAN subnet but keeping it on the lab subnet so all the homelab kit has functioning v6 connectivity. Work has kept me far too busy so haven't extensively tested on that as of yet.

Do let us know what CF say!

Yeah, I was thinking it was mine due to I wasn't using their supplied router, and I already had to phone them to clear the MAC's in their Cache so that my router could connect.
I have to have IPv6 enabled internally due to some of my stuff uses it like NAS Storage etc.

In reply to a post by branflakes:

Do let us know what CF say!

Well its already been about 5 days, but their help desk did say that they was very busy atm, I have also sent them a tweet to show there are issues over IPv6, so I will see what happens.

Paul

In reply to a post by Pheasant:

Crikey you're probably neighbours, just shout out the window! 😀

LOL, well Manor Park is not a small area, but yeah LOL

Paul

In reply to a post by PaulKirby:

In reply to a post by Pheasant:

Crikey you're probably neighbours, just shout out the window! 😀

LOL, well Manor Park is not a small area, but yeah LOL

Paul

I jest but yeah, wonder if its a common network element a router, switch blade etc that's common to you both.

In reply to a post by Pheasant:

In reply to a post by PaulKirby:

In reply to a post by Pheasant:

Crikey you're probably neighbours, just shout out the window! 😀

LOL, well Manor Park is not a small area, but yeah LOL

Paul

I jest but yeah, wonder if its a common network element a router, switch blade etc that's common to you both.

It could very well be just that, I know their lower 127 block of IPv4 Address (masking I am being given) are all GEO Located to Manor Park.

Paul

Looks like they have POP's and peer in Telehouse North and Equinix in the Docklands and Slough. Their speed tester on Ookla always gives me a mighty fine latency result from my London address, probably as my provider has peering in the same datacentre. So its all sub millisecond.

Not sure how far from the POP's they have further distribution gear. Some of their larger grey street cabs may have either mini OLTs or L2/3 switches in them.

TBB single thread reads around 385 today, I've even tried a separate machine to rule out configuration issues. Multi-thread results are consistently above 900.

Trying some speedtest.net single thread gives results in the 300-550Mbit range to London servers.

Hey Branflakes

Is IPv6 is completely dead for you, its completely down for me.

Paul

In reply to a post by PaulKirby:

Hey Branflakes

Is IPv6 is completely dead for you, its completely down for me.

Paul

All good here! Was awake into the small hours working on something - so was working at 1am at least.

Whats your IPv6 BQM like ?

Mine is all red.

Paul

Here's mine (the red is because the firewall was screwing with ICMP and I was redoing some of cabling in the rack) but it's working now. Still have the fringe packet loss however.

In reply to a post by branflakes:

Here's mine (the red is because the firewall was screwing with ICMP and I was redoing some of cabling in the rack) but it's working now. Still have the fringe packet loss however.

I've just set up an IPv6 address and a TBB BQM on it.

I have the same packet loss or a little less in SW16. Router is a Mikrotik CCR1009.

IPv4 is completely clean.

Is this the same for everyone, I wonder?

In reply to a post by ft247:

In reply to a post by branflakes:

Here's mine (the red is because the firewall was screwing with ICMP and I was redoing some of cabling in the rack) but it's working now. Still have the fringe packet loss however.

I've just set up an IPv6 address and a TBB BQM on it.

I have the same packet loss or a little less in SW16. Router is a Mikrotik CCR1009.

IPv4 is completely clean.

Is this the same for everyone, I wonder?

Yeah I think CFL is having bad routing issues on IPv6 because IPv4 is spotless.

*** Ignore the huge red blocks, I killed IPv6 LOL ***
Here are my 2 BQM's IPv6 BQM (MikroTik DDNS) - IPv4 BQM (OWN DDNS) as you can see IPv4 is clean and IPv6 has packet loss at the top.

Well they still haven't phoned back, and I am seeing packets over IPv6 not arriving and I am seeing TCP delays and resends happening.
So that needs to be resolved.

But yeah IPv6 has been like this from day one for me.

Paul

Btw, I have a question to ask, maybe it's not exactly related to this thread.

But is there any difference in performance for Community Fibre in areas with 3Gbps vs areas that have only 1Gbps in the lower speed packages?

I will get Community Fibre in a few months hopefully with 3Gbps max. My other property also has Community Fibre but has 1Gbps max as only option available.

So if you were to choose 1Gbps package in an area that has 3Gbps vs areas that don't have 3Gbps. Are the performances identical or will there be a difference, in speeds and latency? I asked support chat and all they say is that areas with 3Gbps have 10G-XGS PON and those that don't, don't have it.

If this is the case, does it impact performance even in the lower speed packages for those areas that don't have 3Gbps as an option vs those that have?

I wouldn't know our area only has up to 1Gbit Symmetrical Speeds, well the last time I checked.

But even if 3Gbit was in our area I would still go with the 1Gbit speeds due to some of our internal hardware would need to be upgraded and the price for the 3Gbit would be out of what we put aside for broadband.

Paul

In reply to a post by PaulKirby:

I wouldn't know our area only has up to 1Gbit Symmetrical Speeds, well the last time I checked.

But even if 3Gbit was in our area I would still go with the 1Gbit speeds due to some of our internal hardware would need to be upgraded and the price for the 3Gbit would be out of what we put aside for broadband.

Paul

Yeah, that I am fully aware. I know I also wouldn't be taking the 3Gbps even if it became available in my area.

I'm estimated to go live by December, but support can't tell me if it will be 3Gbps yet. Buildings next to me have 3Gbps as an option on 24 month contract, so I'm assuming I will get it too.

The thing I'm curious is whether those areas that are stuck on 1Gbps symmetrical only, have worse speed performance and latency vs 1Gbps in those areas that have 3Gbps as an option.

Maybe it won't show much differences. But if indeed 3Gbps has XGS-PON in such areas. This is something that has always been bugging my mind.

This leads me to believe that certain areas have only GPON even if it also does 1Gbps symmetrical. My concern is whether GPON would be more prone to crosstalk/contention with more customers signing up to a gigabit package versus those on XGS-PON yielding to reduced speeds over time.

In reply to a post by BLaZiNgSPEED:

The thing I'm curious is whether those areas that are stuck on 1Gbps symmetrical only, have worse speed performance and latency vs 1Gbps in those areas that have 3Gbps as an option.

Maybe it won't show much differences. But if indeed 3Gbps has XGS-PON in such areas. This is something that has always been bugging my mind.

This leads me to believe that certain areas have only GPON even if it also does 1Gbps symmetrical. My concern is whether GPON would be more prone to crosstalk/contention with more customers signing up to a gigabit package versus those on XGS-PON yielding to reduced speeds over time.

If you take the 1Gb/s service in an area with XGS-PON available they may still give you a GPON ONT.

You would only need an XGS-PON ONT if you took the 3Gb/s service.
If a PON was looking busy they could move customers to XGS-PON, but imo that's unlikely to be an issue for the majority.

Perhaps someone with 3Gb/s available but who takes 1Gb/s could confirm.

Edit: confirmed below, everything built going forward is XGS-PON.

Edited by j0hn83 (Thu 05-Aug-21 09:07:44)

My area is very recent expansion for CF, and borders an area that has 3Gbit available.

My area does not have 3Gbit for sale, which is the same situation as another new CF build area I'm aware of.

I have the 1Gbit service and an XGS-PON ONT.

To me, this means there is another reason for restricting 3Gbit sale - either commercial, or limited backhaul.

Apparently the CEO in this November article confirmed that only half the network was able to handle the 3 Gbps direct connections, despite all the new network being constructed as XGS-PON.

https://www.capacitymedia.com/articles/3826840/commu...

They close the article with; ‘The company is connected to two data centres in east London, Telehouse North and Equinix London LD8, and its network runs though 80 of its own cabinets as well as “a few BT exchanges”.’

Perhaps it’s indeed a backhaul limitation to those cabs and exchanges rather than the PON.

In reply to a post by Pheasant:

How are you getting on with the service? Did you ever get to the bottom of the difference you saw with TBB single-thread DL performance vs multi-thread?

Well, I've got to the bottom of it - I need a better router.

I'm using an old Mikrotik CCR1009 which was removed from a customer due to damaged Ethernet ports. The SFP+ port works fine and is all I need, so sounds ideal.

Bypassing the router gets me 850+ single thread down on TBB. Using the router and cleaning up the firewall rules gets me to 380 single thread (TBB) and 550 single thread (Ookla).

Another issue is that when using the CCR1009 as an OpenVPN endpoint, throughput is <100Mbit. In general the Tilera processor models are not great at single thread tasks.

Now the question is what to replace it with!

I’m replacing mine with 2004’s. Much better VPN and single thread performance.

I use a MikroTik Tik RB4011iGS+RM as the router here which works great.

Paul

Well I had Community Fibre do checks and they said no packets are being dropped, even though there are due to I played World of Warcraft over IPv6 and I either saw lag (prob due to resends) or my connection was completely dropped.

Their last response:
Hello Paul,

Our Networking team have conducted thorough checks and have run tests from the traceroute provided by client which showed on Hop 8 that there is 0% packet loss and this specific hop is beyond our ASN.
We also tried from the Client Gateway Router and it was still showing no packet loss whatsoever.
They have advised there is nothing wrong which can be seen on their end after thorough testing, and have advised for you to monitor it accordingly.
They have suggested potentially this could be occurring at a specific time so if there is any further information you can provide then that would also be helpful.

Sincerely,
** name removed **
Technical Support Agent
Community Fibre Ltd.

So not sure what I can do, but it couldn't of been that thorough due to all my checks are showing it.

Paul

What’s your IPv6 BQM like at the moment? Can you setup another BQM to the next hop router to yours?

Don't forget that you aren't detecting the lost pings. thinkbroadband is of course. So the problem may not be your connection, but anywhere on the IPv6 routing from them to and back from you.

Edited by pluralist (Mon 16-Aug-21 09:59:45)

Paul - I got something very similar from CF. Their reasoning was that the intermediary routers in their network deprioritise ICMP, therefore don't always respond to pings. Surely if this was the case, the same would be seen in the v4 BQM, which it isn't.

Here's the BQM to the assigned interface address

and here's the next hop, 2a02:6b60:0:82::1 (possibly OLT or first switch?) - it's newly set up from now, so not populated. You may have the same considering we're in the same area and likely same /25.

Edited by branflakes (Mon 16-Aug-21 11:04:47)

Well strange thing is Pings are fine, sadly whatever the IPv6 address I route to has issues.

It is possible that some of the hops are not prioritising the packets when doing the trace routes, but pings are still working without any issues at all.

My IPv6 BQM
Next Hop BQM: 2a02:6b60:0:84::1 Yes I am seeing no packet loss stuff at the top.

I did a route to google.com and got this:

/tool traceroute 2a00:1450:4009:818::200e
 # ADDRESS                          LOSS SENT    LAST     AVG    BEST   WORST STD-DEV STATUS
 1 2a02:6b60:0:84::1                 97%   34 timeout     1.7     1.7     1.7       0
 2 2a02:6b60:0:1:4::59               97%   33 timeout     1.9     1.9     1.9       0
 3 2a02:6b60:0:1:1::59               97%   33 timeout     2.1     2.1     2.1       0
 4 2a02:6b60:0:1:2::55               97%   33 timeout     1.7     1.7     1.7       0
 5 2a02:6b60:0:1:3::62                0%   33   1.5ms     1.7     0.7       3     0.5
 6 2001:4860::1c:4001:1ad7          72..   33 timeout      43     2.2   363.9   113.5
 7 2001:4860:0:1::2fd9              51..   33 timeout     3.2     2.8     3.5     0.2
 8 2a00:1450:4009:818::200e           0%   33   2.3ms     2.2       2     2.5     0.1

Hop 1 is the first hop my router takes.

Paul

Edited by PaulKirby (Mon 16-Aug-21 13:06:45)

I would approach CFL jointly, as your both seeing similar levels of ipv6 packet loss, yet are on different next hop routers / devices suggesting there is indeed something common / deeper within the CFL network that is causing the sustained packet loss that you are both seeing.

Good idea. I know ft247 also reported seeing the same BQM on the v6 interface address and is on the other side of London from both Paul and me, so perhaps also worth them chiming in on zero PL on the hop just before our routers to show it's not just in one area.

In reply to a post by Pheasant:

I’m replacing mine with 2004’s. Much better VPN and single thread performance.

That's what I've just done. The 2004-1G-12S+2XS seems much better suited to the task at hand, even with substantially the same busy config as the 1009.

Happy to report that single thread performance is now essentially the same as multithread - a flat line at 940Mbit all day. There may even be an approx 0.05msec latency improvement too.

The Whitebox is reporting constant 940Mbit upload and download, both single and multi-thread.

TBB and Ookla concur on download, 940Mbit flat line single and multi-thread. On upload TBB bounces around a bit more in download but comfortably achieves 650-850Mbit. Ookla is usually 800+.

There is still more to look at on upload speed to distant servers - there are a number of cases where the distant server will give an easy 940Mbit download but struggle to receive more than 300 from me.

Sounds good. I'm still procrastinating the move to the 2004's. The CCR10's are carrying on for a bit more.

How are you getting on with ROS 7.x ?

In reply to a post by ft247:

There is still more to look at on upload speed to distant servers - there are a number of cases where the distant server will give an easy 940Mbit download but struggle to receive more than 300 from me.

How "distant" in terms of milliseconds RTT?

If this is a single threaded upload, then it's almost certainly due to a tiny amount of packet loss somewhere along the line limiting TCP throughput.

https://en.wikipedia.org/wiki/TCP_tuning#Packet_loss

In reply to a post by Pheasant:

How are you getting on with ROS 7.x ?

Mine is the SFP+ model and running 6.48.4 - it's only the mostly copper 2004 which is 7.x only.

I'm getting close to 'happy to try v7' territory, but need a bit more free time on my hands before diving in.

In reply to a post by candlerb:

How "distant" in terms of milliseconds RTT?

If this is a single threaded upload, then it's almost certainly due to a tiny amount of packet loss somewhere along the line limiting TCP throughput.

Less than 50ms - in many cases less than 30. I think it starts appearing around the Scotland/Germany level of distance from London. Will try to post back soon with more detail.

In reply to a post by ft247:

Less than 50ms - in many cases less than 30. I think it starts appearing around the Scotland/Germany level of distance from London. Will try to post back soon with more detail.

At 20ms RTT, a single TCP stream will be limited to ~300Mbps by packet loss of just .0004%, or 1 in 250,000 packets

1460/(.02*sqrt(.000004))*8 = 292,000,000

In reply to a post by ft247:

My area is very recent expansion for CF, and borders an area that has 3Gbit available.

My area does not have 3Gbit for sale, which is the same situation as another new CF build area I'm aware of.

I have the 1Gbit service and an XGS-PON ONT.

To me, this means there is another reason for restricting 3Gbit sale - either commercial, or limited backhaul.

Community Fibre just went live in my area 2 days ago and I can confirm 3Gbit is available to order! So indeed it is XGS-PON that is being used.

One thing that is also interesting is that now my other 2 apartments one in Shoreditch and the other one in Sidney Street also now show up 3Gbps as available to order when previously they would only show as 1Gbps available.

It is quite possible that if you now check the website for your postcode you will also see 3Gbit available for sale if you select the 24 month contract package.

The fact that they all show 3Gbps available suggests that indeed they all had XGS-PON ONT but simply as you say the backhaul limitation was probably something that needed addressing.

Anyway, now I have to wait for my TalkTalk FTTC contract to end in February 2022 and then I'll switch to Community Fibre. The new service will surely be a much better service than FTTC even if XGS-PON wasn't available. But now it is much more reassuring!

Sorry to bump this old thread but I have the same issue.

Here's the current TBB graph:

My Broadband Ping

Ignore the big red blocks, that's just me fiddling/trying to find the correct IPv6 address to ping (apparently my public IPv6 address is on 2nd hop, not 1st hop).

There's consistent packet loss between 0% and 5% that's only seen in IPv6. IPv4 has no packet loss at all.

I also ran tracert (via WinMTR) and the below is what I found out so far:

IPv6
IPv4

As WinMTR constantly sends out loads of pings, I expect some loss to some degree as result of servers not responding to some of my constant pings. The %% (meaning packet loss) as observed in IPv6 result is huge compared to IPv4 result.

I've looked through the entire thread, it seems like nobody has found a solution yet? Or did I misread something?

I cannot speak for the service yet. When I previously posted in October 2021 my Community Fibre service didn't go live until December 2022!

So, due to this uncertainty at that time and the fact that TalkTalk FTTC price hiked my service to £40 a month, I switched to BT.

Here I am still on BT FTTC and am generally happy with their service. I have not joined Community Fibre yet as I said I was because I haven't yet found a good deal. The good deals that were on offer were available when the service wasn't yet live in my building.

It seems the packet losses are reported by other users as well. I'm assuming that the problem was resolved but came back again. Hopefully it will be fixed again. Are you testing this with WiFi or Ethernet cable?

I'm reading that this problem is being experienced for IPv6 but not IPv4. Also, I read that Community Fibre have a CGNAT restriction on all packages except for the 3Gbps so not sure if that also has an impact on packet loss and ping performance.

I'm seeing a similar IPv6 packet loss - here is the link to my graph

Unfortunately I can't monitor IPv4 as I am behind CGNAT.

I used to get the same thing when I used monitor the IPv6 Address of the router, yet the Routers Gateway has no packet loss / red marks.
It got to a point I was checking a lot even though there was no actual issues on the connection, so I just disabled them and just monitor my IPv4 Address (shown in my Sig).

But I have just temporary (for the next couple of days) restarted up my IPv6 Monitors shown below:
My Routers IPv6 Connection: Router IPv6 Connection
As well as the first hope out (i.e. Routers IPv6 Gateway) First IPv6 Hop out / Routers Gateway

As you will see the routers IPv6 Monitor will have the little RED spikes on top where as the first hop out (router's gateway) is fine, so something is up.

Also traceroute from my PC shows timeouts on some of the Hops, that's could still be fine, not all hops respond to pings.

I know when I last looked at this a couple of years ago I did speak to somebody at CFL and they couldn't see any packet loss or any issues that would cause this.
I know I am not seeing or having any issues over IPv6 and I use it a lot.

tracert google.co.uk

Tracing route to google.co.uk [2a00:1450:4009:826::2003]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  [ROUTER IPv6]
  2    <1 ms     1 ms    <1 ms  [ISP GATEWAY]
  3     2 ms     1 ms     1 ms  lag-6.acc1.med.lon.network.as201838.net [2a02:6b60:0:1:4::59]
  4     *        1 ms     2 ms  lag-17.edge1.thn.lon.network.as201838.net [2a02:6b60:0:1:4::170]
  5     2 ms     1 ms     1 ms  2001:4860:1:1::814
  6     2 ms     2 ms     2 ms  2001:4860:0:1::7e3f
  7     2 ms     2 ms     2 ms  2001:4860:0:1::73c5
  8     2 ms     1 ms     1 ms  lhr48s48-in-x03.1e100.net [2a00:1450:4009:826::2003]

Trace complete.

The as201838 are the Community Fibre Hops and those are the ones that I am having issues with that are timeouts due to Traceroute on my router does get a response now and then and the second as201838 (lag-17.edge1.thn.lon.network.as201838.net [2a02:6b60:0:1:4::170]) that has the highest packet loss.

But as I said I am not seeing or having any issues over IPv6.

---
Paul

Thanks, Paul

When running WinMTR I get packet loss after the ISP Gateway and going to 2a02:6b60:0:1::10 which you point out is still within CFL.

I will continue to monitor for now, but as you say, everything seems to work OK.

Yeah, no point worrying about it if you are not having any issues.
I have now disabled those BQM's and just left my IPv4 one running like I have always done.

But I created a Snapshot of them linked below:
My Routers IPv6 Connection BQM - My Routers IPv6 Gateway (Next Hop Out) BQM

But yeah like I said those hops don't always respond to pings, be nice if they all did, but not all do.

As to doing a tracerote, they are hit and miss when over IPv6 so to get a complete route I had to run several times, or just leave my router doing them for like 5 mins.

But yeah, I wouldn't worry about it until you have actual issues.

---
Paul

Thank you. Seems like a non-issue.

Just out of curiosity, since you mentioned you use MikroTik in your signature, do you notice any improvements compared to Linksys?

In reply to a post by rtcw:

Thank you. Seems like a non-issue.

Just out of curiosity, since you mentioned you use MikroTik in your signature, do you notice any improvements compared to Linksys?

I never tried the Linksys Router myself (not a fan of using ISP Supplied Routers), I know the installers set everything up and did the tests using it and commented my hardware was better than what they installed and then took a photo of the Speedtest results with my hardware.

With my MikroTik RB4011iGS+ Router I already hit just over the max limit of the 1GBit limit (without Jumbo Frames) so I won't really see any performance increase unless CFL changes the AdTran Modem to a model that has a 10Gbit Ethernet or SFP+ Port and I cannot see that happening unless I upgrade to their 3 Gbit Package, even though the price for the 3Gbit Pakage is the same as BT 900:110 Package + Phone + Line Rental.
So the budget is there, but at this point of time we don't need that speed.

I am in the process of upgrading our LAN to 10Gbit LAN where I will be using a MikroTik CCR2004-1G-12S+2XS I know its overkill, but it just means it will last longer.

---
Paul